Abstract: SA-PO755
Underperformance of ESRD Prediction Equations in a Contemporary Diverse Adult Population with Advanced CKD
Session Information
- CKD: Epidemiology, Risk Factors, Prevention - III
October 27, 2018 | Location: Exhibit Hall, San Diego Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: CKD (Non-Dialysis)
- 1901 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention
Authors
- Zheng, Sijie, The Permanente Medical Group, Oakland, California, United States
- Pravoverov, Leonid, Kaiser Permanente, Walnut Creek, California, United States
- Shiau, Victor, Kaiser Permanente Northern California, Oakland, California, United States
- Parikh, Rishi V., Kaiser Permanente, Walnut Creek, California, United States
- Tan, Nhi, Kaiser Oakland Medical Center, OAKLAND, California, United States
- Mroz, Joanna, The Permanente Medical Group, Oakland, California, United States
- Tan, Thida C., Kaiser Permanente Northern California, Oakland, California, United States
- Jonelis, Tracy Y., The Permanente Medical Group, Inc, San Francisco, California, United States
- Go, Alan S., Kaiser Permanente Northern California, Oakland, California, United States
Background
Two risk prediction models (Tangri, KPNW) have been proposed for risk stratification for transition to ESRD in adults with CKD 3-5. We evaluated the performance of these models in an ethnically diverse advanced CKD population.
Methods
We identified all adult members of Kaiser Permanente Northern California (KPNC) with Stage 4-5 CKD between Jan. 2008-Sep. 2015. Receipt of RRT through Sep. 2016 was ascertained from an internal ESRD registry. Data on patient characteristics and lab results were obtained from electronic health records. We examined model discrimination (c statistic) and calibration (observed vs. predicted risk) for initiating RRT at 1, 2 and 5 years for the Tangri equation and at 2 years for the KPNW equation.
Results
In 17,586 adults with CKD 4-5, mean age was 73.7 years, 53% were women, and 49% were persons of color. The 1-, 2- and 5-year crude risks for RRT were 8.8% (95%CI:8.3-9.2%), 20.1% (19.5-20.8%) and 37.9% (36.9-38.8%), respectively. For Tangri, we found lower discrimination at 1 year (c=0.80 KPNC, c=0.86 Tangri), and 5 years (c=0.76 KPNC, c=0.84 Tangri), with substantial underestimation of actual absolute risk (Figure 1A). Similarly, for the KPNW model, we observed lower discrimination (c=0.79 KPNC, c=0.96 KPNW) and underestimation of actual risk (Figure 1B).
Conclusion
In a large, ethnically diverse CKD 4-5 population, Tangri and KPNW prediction models exhibited both lower discrimination and significant underestimation of absolute RRT risk. Given that most ESRD cases are derived from CKD 4-5 patients, more accurate prediction models are needed to help tailor management.
Observed vs. Predicted Risk of RRT at 2 years by Deciles of Predicted Risk in Stage 4/5 CKD Patients at Kaiser Permanente Northern California (Jan. 2008-Sep. 2015).
Funding
- Private Foundation Support